The LM-Cut Heuristic Family for Optimal Numeric Planning with Simple Conditions

Authors: Ryo Kuroiwa, Alexander Shleyfman, Chiara Piacentini, Margarita P. Castro, J. Christopher Beck

JAIR 2022 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our extensive empirical evaluation shows that the new LM-cut heuristic, both on its own and as part of the operator counting framework, is the state-of-the-art for this class of numeric planning problem.
Researcher Affiliation Collaboration Ryo Kuroiwa EMAIL Department of Mechanical and Industrial Engineering University of Toronto Canada Alexander Shleyfman EMAIL The Department of Computer Science Bar-Ilan University Israel Chiara Piacentini EMAIL Augmenta Inc Canada Margarita P. Castro EMAIL Department of Industrial Engineering and Systems Pontificia Universidad Cat olica de Chile Chile J. Christopher Beck EMAIL Department of Mechanical and Industrial Engineering University of Toronto Canada
Pseudocode Yes The pseudo-code is presented in Appendix C.
Open Source Code Yes We implemented the heuristics in Numeric Fast Downward (NFD) (Aldinger & Nebel, 2017)4 using C++11 with GCC 7.5.0 on Ubuntu 18.04. 4. https://github.com/Kurorororo/numeric-fast-downward
Open Datasets Yes We consider domains with simple conditions from the literature (Scala et al., 2016, 2017, 2020).
Dataset Splits No The paper mentions different planning domains and instances (e.g., Small Counters (8), Counters (8), Farmland (30)), and discusses excluding certain instances or adding satisficing versions. This describes the selection of problem instances from established benchmarks rather than explicit training/validation/test splits of a single dataset. Standard dataset splits in the machine learning sense are not provided.
Hardware Specification Yes In all the experiments, we evaluate the heuristics inside an A search imposing a 30 minute time limit and 4 GB memory limit on an Intel(R) Xeon(R) CPU E5-2620 @ 2.00GHz processor.
Software Dependencies Yes We implemented the heuristics in Numeric Fast Downward (NFD) (Aldinger & Nebel, 2017)4 using C++11 with GCC 7.5.0 on Ubuntu 18.04... the mathematical programming solver used for these heuristics is IBM CPLEX 12.10.
Experiment Setup Yes In all the experiments, we evaluate the heuristics inside an A search imposing a 30 minute time limit and 4 GB memory limit... For the numeric heuristics, we add the redundant constraints to the goal conditions and preconditions of actions in the same fashion as Scala et al. (2016a).